fix(fuse,core): what a create sent, and where, outlives a restart and a rename (B182, B183, B184)

The known-bug confirmation run failed one_client_on_a_flaky_link seeds 6
and 11 on three bugs around a create whose answer was lost:

- B182: after a restart, a write over a file whose create landed but was
  never read back forked a conflict copy. A revision of a file that has
  no revision of Drive's now adopts the landed node first.
- B183: the SHA-1s a create sent were kept in memory, so after a restart
  its trash or retry no longer knew the node Drive made. They are now on
  the op's row (pending_op.sent_sha1, schema 42).
- B184: a rename rewrote the op's target, and the trash and the retry
  looked for the node under the last name only. Each attempt's parent
  and name are recorded (pending_op.sent_to); the trash looks under all
  of them, and a retry adopts the node an earlier attempt made and moves
  it to the current name.

crates/pdfs-core/src/db/migrations.rs

@@ -11,7 +11,7 @@ use super::Db;
 use crate::Result;
 
 /// Current schema version. Bump on every forward migration added below.
-pub(super) const SCHEMA_VERSION: i64 = 41;
+pub(super) const SCHEMA_VERSION: i64 = 42;
 
 impl Db {
     pub(super) fn migrate(&self) -> Result<()> {
@@ -424,6 +424,22 @@ impl Db {
                 tx.execute_batch(MIGRATION_V41)?;
             }
         }
+        if current < 42 {
+            // Same guards as V26-V40.
+            let has_column: bool = tx.query_row(
+                "SELECT COUNT(*) FROM pragma_table_info('pending_op') WHERE name = 'sent_sha1'",
+                [],
+                |row| row.get::<_, i64>(0),
+            )? > 0;
+            let has_ops: bool = tx.query_row(
+                "SELECT COUNT(*) FROM sqlite_master WHERE type = 'table' AND name = 'pending_op'",
+                [],
+                |row| row.get::<_, i64>(0),
+            )? > 0;
+            if has_ops && !has_column {
+                tx.execute_batch(MIGRATION_V42)?;
+            }
+        }
         tx.execute(
             "INSERT INTO sync_state (key, value) VALUES ('schema_version', ?1)
              ON CONFLICT(key) DO UPDATE SET value = excluded.value",
@@ -1302,3 +1318,15 @@ UPDATE nodes
 UPDATE nodes SET uid = 'local~' || lid WHERE lid IN (SELECT lid FROM local_v41);
 DROP TABLE local_v41;
 ";
+
+/// Schema v42: the SHA-1 of each blob a queued create sent, space-separated.
+/// A create whose answer was lost may have made its file with any of them,
+/// and the op that finds that file again may run after a restart. `NULL` for
+/// an op that has sent nothing, which is every op migrated: what the daemon
+/// that wrote it sent is not known. Likewise the parent and name each attempt
+/// was sent under, as a JSON list of pairs: a rename since then moves the op
+/// but not the file Drive may have made.
+const MIGRATION_V42: &str = "
+ALTER TABLE pending_op ADD COLUMN sent_sha1 TEXT;
+ALTER TABLE pending_op ADD COLUMN sent_to TEXT;
+";

crates/pdfs-core/src/db/ops.rs

@@ -765,6 +765,89 @@ impl Db {
         })
     }
 
+    /// Record that queued create `id` is sending a blob with SHA-1 `sha`.
+    ///
+    /// Drive may make the file and lose its answer. The op that next finds
+    /// the file, the create's retry or the trash a delete made of it, takes
+    /// it for its own when it holds any bytes sent so. Kept on the row, so a
+    /// newer blob, the turn into a trash and a restart all keep it
+    /// (`docs/BUGS.md` B156, B183).
+    pub fn note_create_sent(&self, id: i64, sha: &str) -> Result<()> {
+        self.conn.lock().execute(
+            "UPDATE pending_op
+             SET sent_sha1 = CASE
+                 WHEN sent_sha1 IS NULL THEN ?2
+                 WHEN instr(' ' || sent_sha1 || ' ', ' ' || ?2 || ' ') > 0 THEN sent_sha1
+                 ELSE sent_sha1 || ' ' || ?2 END
+             WHERE id = ?1",
+            params![id, sha],
+        )?;
+        Ok(())
+    }
+
+    /// Record that queued create `id` is being sent to `parent` under `name`.
+    ///
+    /// A rename while the op waits rewrites its target, but the file a lost
+    /// answer left stays where it was sent. The retry and the trash look
+    /// there too (`docs/BUGS.md` B184).
+    pub fn note_create_target(&self, id: i64, parent: &str, name: &str) -> Result<()> {
+        let mut conn = self.conn.lock();
+        let tx = conn.transaction()?;
+        let sent: Option<String> = tx
+            .query_row("SELECT sent_to FROM pending_op WHERE id = ?1", [id], |r| {
+                r.get(0)
+            })
+            .optional()?
+            .flatten();
+        let mut targets: Vec<(String, String)> = sent
+            .and_then(|json| serde_json::from_str(&json).ok())
+            .unwrap_or_default();
+        let target = (parent.to_string(), name.to_string());
+        if !targets.contains(&target) {
+            targets.push(target);
+            tx.execute(
+                "UPDATE pending_op SET sent_to = ?2 WHERE id = ?1",
+                params![id, serde_json::to_string(&targets)?],
+            )?;
+        }
+        tx.commit()?;
+        Ok(())
+    }
+
+    /// The parent and name queued op `id` was sent under as a create, in the
+    /// order it was ([`Db::note_create_target`]).
+    pub fn create_targets(&self, id: i64) -> Result<Vec<(String, String)>> {
+        let sent: Option<String> = self
+            .conn
+            .lock()
+            .query_row("SELECT sent_to FROM pending_op WHERE id = ?1", [id], |r| {
+                r.get(0)
+            })
+            .optional()?
+            .flatten();
+        Ok(sent
+            .and_then(|json| serde_json::from_str(&json).ok())
+            .unwrap_or_default())
+    }
+
+    /// The SHA-1 of each blob queued op `id` sent as a create
+    /// ([`Db::note_create_sent`]).
+    pub fn create_sent(&self, id: i64) -> Result<Vec<String>> {
+        let sent: Option<String> = self
+            .conn
+            .lock()
+            .query_row(
+                "SELECT sent_sha1 FROM pending_op WHERE id = ?1",
+                [id],
+                |r| r.get(0),
+            )
+            .optional()?
+            .flatten();
+        Ok(sent
+            .map(|sent| sent.split(' ').map(str::to_string).collect())
+            .unwrap_or_default())
+    }
+
     /// Replace the sidecar of a queued op, for a baseline that has moved under
     /// it.
     ///

crates/pdfs-core/src/db/tests.rs

@@ -3848,6 +3848,69 @@ fn a_create_withdrawn_after_it_failed_queues_a_trash_that_keeps_its_blob() {
     assert!(db.pending_ops().unwrap().is_empty());
 }
 
+/// What a queued create sent, and where, stays on its row: through a newer
+/// blob, through the turn into a trash a delete makes of it, and across a
+/// reopen. The trash finds the file the create made by those bytes, after a
+/// restart too, and under a name the file had before a rename
+/// (`docs/BUGS.md` B183, B184).
+#[test]
+fn what_a_create_sent_outlives_a_newer_blob_its_trash_and_a_restart() {
+    let path = std::env::temp_dir().join(format!(
+        "pdfs-create-sent-{}-{}.db",
+        std::process::id(),
+        now_test_id()
+    ));
+    let parent = uid("parent").to_string();
+    let id = {
+        let db = Db::open(&path).unwrap();
+        let (id, _) = db
+            .enqueue_op(&PendingOp {
+                id: 0,
+                kind: OP_CREATE.to_string(),
+                uid: "local~sent".to_string(),
+                parent_uid: Some(parent.clone()),
+                name: Some("d".to_string()),
+                blob_path: Some("/staging/first".to_string()),
+                meta_json: Some("{}".to_string()),
+                created_at: 1,
+                attempts: 0,
+                last_error: None,
+                next_attempt_at: 0,
+            })
+            .unwrap();
+        assert!(db.create_sent(id).unwrap().is_empty());
+        assert!(db.create_targets(id).unwrap().is_empty());
+        db.note_create_target(id, &parent, "c").unwrap();
+        db.note_create_target(id, &parent, "c").unwrap();
+        db.note_create_sent(id, "aaaa").unwrap();
+        db.note_create_sent(id, "aaaa").unwrap();
+        db.record_op_failure(id, "no answer", 0).unwrap();
+        db.attach_blob_to_create("local~sent", "/staging/second", "{}")
+            .unwrap()
+            .expect("the create is still queued");
+        db.note_create_target(id, &parent, "d").unwrap();
+        db.note_create_sent(id, "bbbb").unwrap();
+        db.record_op_failure(id, "no answer", 0).unwrap();
+        db.delete_ops_for_uid("local~sent").unwrap();
+        id
+    };
+
+    let db = Db::open(&path).unwrap();
+    let ops = db.pending_ops().unwrap();
+    assert_eq!(ops.len(), 1);
+    assert_eq!((ops[0].id, ops[0].kind.as_str()), (id, OP_TRASH));
+    assert_eq!(db.create_sent(id).unwrap(), vec!["aaaa", "bbbb"]);
+    assert_eq!(
+        db.create_targets(id).unwrap(),
+        vec![
+            (parent.clone(), "c".to_string()),
+            (parent.clone(), "d".to_string())
+        ]
+    );
+    drop(db);
+    let _ = std::fs::remove_file(path);
+}
+
 /// A create deleted after it was sent becomes a trash in its place. The
 /// create's backoff, and the failure of an attempt still on the wire, are the
 /// create's: the trash is due at once (`docs/BUGS.md` B171).

crates/pdfs-fuse/src/drain.rs

@@ -153,6 +153,19 @@ enum DrainDisposition {
     AuthorityUnknown(NodeUid),
 }
 
+/// The node an earlier attempt at a create made before a rename moved the op
+/// (`docs/BUGS.md` B184).
+struct EarlierTwin {
+    /// The folder the attempt was sent to.
+    parent: NodeUid,
+    /// The name it was sent under.
+    name: String,
+    /// The node it made.
+    twin: NodeUid,
+    /// Whether the node holds the op's blob already.
+    holds_blob: bool,
+}
+
 /// Why an op went back on the clock without consuming an attempt. Only affects
 /// what the user is told — every variant is deferred and reported identically.
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
@@ -558,7 +571,6 @@ impl Core {
                 self.release_claim(op);
                 return;
             }
-            self.create_sent.lock().remove(&op.id);
             return;
         }
         match fetched {
@@ -747,8 +759,13 @@ impl Core {
         let Some(real) = parse_node_uid(&op.uid) else {
             return Ok(());
         };
-        let Some(local) = self.unadopted.lock().get(&real).cloned() else {
+        let local = match self.unadopted.lock().get(&real).cloned() {
-            return Ok(());
+            Some(local) => local,
+            // The map does not outlive a restart, which also forgets the
+            // stand-in: the row is under the real uid by then. A file as it was
+            // made here has no revision of Drive's (`docs/BUGS.md` B182).
+            None if op.kind == OP_REVISION && self.never_read_back(&real) => real.clone(),
+            None => return Ok(()),
         };
         // A node gone since is the op's to deal with.
         if self.fetch_node_remote(&real)?.is_some() {
@@ -759,6 +776,16 @@ impl Core {
         Ok(())
     }
 
+    /// Whether the tree has the file `real` as it was made here: its create
+    /// landed, but Drive's answer was never read back.
+    fn never_read_back(&self, real: &NodeUid) -> bool {
+        let st = self.state();
+        st.by_uid
+            .get(real)
+            .and_then(|ino| st.entries.get(ino))
+            .is_some_and(|entry| !entry.node.is_folder() && node_revision_id(&entry.node).is_none())
+    }
+
     /// Apply a queued rename/move to the remote.
     ///
     /// The op is the desired end state, so the remote's current state decides
@@ -953,7 +980,6 @@ impl Core {
                 None => debug!(%uid, name, "withdrawn create never landed; trash op satisfied"),
             }
             self.retire_trash_op(op, &uid)?;
-            self.create_sent.lock().remove(&op.id);
             return Ok(());
         }
         match self
@@ -989,24 +1015,48 @@ impl Core {
     }
 
     /// The node a create that a trash withdrew made on Drive although its
-    /// answer was lost: in the create's folder under its name, and one the
+    /// answer was lost: in the create's folder under its name, or under one
-    /// create would have adopted had it been retried ([`adoptable`]). A node
+    /// an attempt was sent with before a rename (`docs/BUGS.md` B184).
-    /// with a change of ours queued is one the user is still working on.
     fn withdrawn_twin(
         &self,
         op: &PendingOp,
     ) -> Result<Option<NodeUid>, Box<dyn std::error::Error>> {
-        let (Some(parent), Some(name)) = (
+        let mut targets: Vec<(String, String)> = op
-            op.parent_uid
+            .parent_uid
-                .as_deref()
+            .clone()
-                .filter(|parent| !is_local_uid_str(parent))
+            .zip(op.name.clone())
-                .and_then(parse_node_uid),
+            .into_iter()
-            op.name.as_deref(),
+            .collect();
-        ) else {
+        for target in self.db.create_targets(op.id)? {
-            return Ok(None);
+            if !targets.contains(&target) {
-        };
+                targets.push(target);
+            }
+        }
+        for (parent, name) in targets {
+            if is_local_uid_str(&parent) {
+                continue;
+            }
+            let Some(parent) = parse_node_uid(&parent) else {
+                continue;
+            };
+            if let Some(twin) = self.withdrawn_twin_in(op, &parent, &name)? {
+                return Ok(Some(twin));
+            }
+        }
+        Ok(None)
+    }
+
+    /// The node in `parent` under `name` a withdrawn create made, if it is one
+    /// the create would have adopted had it been retried ([`adoptable`]). A
+    /// node with a change of ours queued is one the user is still working on.
+    fn withdrawn_twin_in(
+        &self,
+        op: &PendingOp,
+        parent: &NodeUid,
+        name: &str,
+    ) -> Result<Option<NodeUid>, Box<dyn std::error::Error>> {
         let uids =
-            match self.block_on_bounded(self.drive.enumerate_folder_children_node_uids(&parent)) {
+            match self.block_on_bounded(self.drive.enumerate_folder_children_node_uids(parent)) {
                 Ok(uids) => uids,
                 Err(e) if is_gone(&e) => return Ok(None),
                 Err(e) => return Err(e.into()),
@@ -1090,11 +1140,6 @@ impl Core {
         }
         let created = self.create_local_node(op);
         self.creating.lock().remove(&op.id);
-        // A withdrawn create goes on as a trash under the same id, which still
-        // looks for the bytes it sent (`Core::withdrawn_twin`).
-        if !self.db.op_exists(op.id).unwrap_or(true) {
-            self.create_sent.lock().remove(&op.id);
-        }
         created
     }
 
@@ -1123,8 +1168,27 @@ impl Core {
         // whose trash is still queued is about to let go of it.
         let mut name = wanted.clone();
         let mut home = parent.clone();
+        let mut sent = (parent_str.to_string(), wanted.clone());
         let mut blob_landed = true;
-        let mut real = self.create_drained_node(op, &parent, &name);
+        // An earlier attempt, sent before the file was renamed or moved, may
+        // have made it and lost the answer. That node is ours where it was
+        // sent, and the rename follows it (`docs/BUGS.md` B184).
+        let mut real = match self.earlier_twin(op, parent_str, &wanted)? {
+            Some(EarlierTwin {
+                parent: was_in,
+                name: was_named,
+                twin,
+                holds_blob,
+            }) => {
+                info!(%local, %twin, was_named, wanted, "an earlier attempt landed unanswered under another name; adopting it");
+                sent = (was_in.to_string(), was_named.clone());
+                home = was_in;
+                name = was_named;
+                blob_landed = holds_blob;
+                Ok(twin)
+            }
+            None => self.create_drained_node(op, &parent, &name),
+        };
         if real.as_ref().is_err_and(|e| is_already_exists(e.as_ref()))
             && let Some((twin, holds_blob)) = self.adoptable_twin(op, &parent, &wanted)?
         {
@@ -1202,7 +1266,7 @@ impl Core {
         let landing = CreateLanding {
             local: &local.to_string(),
             real: &real.to_string(),
-            sent: (parent_str, &wanted),
+            sent: (&sent.0, &sent.1),
             landed: (&home.to_string(), &name),
         };
         let newer = match self.retire_create(op, uploaded, &landing, &local, &real)? {
@@ -1335,6 +1399,38 @@ impl Core {
         Ok(())
     }
 
+    /// The node an earlier attempt at create `op` made under a parent or name
+    /// the op has been moved from since ([`Core::adoptable_twin`]).
+    fn earlier_twin(
+        &self,
+        op: &PendingOp,
+        parent: &str,
+        name: &str,
+    ) -> Result<Option<EarlierTwin>, Box<dyn std::error::Error>> {
+        for (was_in, was_named) in self.db.create_targets(op.id)? {
+            if (was_in.as_str(), was_named.as_str()) == (parent, name) {
+                continue;
+            }
+            let Some(was_in) = parse_node_uid(&was_in) else {
+                continue;
+            };
+            match self.adoptable_twin(op, &was_in, &was_named) {
+                Ok(Some((twin, holds_blob))) => {
+                    return Ok(Some(EarlierTwin {
+                        parent: was_in,
+                        name: was_named,
+                        twin,
+                        holds_blob,
+                    }));
+                }
+                Ok(None) => {}
+                Err(e) if is_gone(e.as_ref()) => {}
+                Err(e) => return Err(e),
+            }
+        }
+        Ok(None)
+    }
+
     /// The node already holding `name` under `parent`, if it is the one an
     /// earlier attempt at this op made before its answer was lost, and whether
     /// it holds the op's blob already.
@@ -1407,12 +1503,10 @@ impl Core {
     /// `current`, those of the blob it holds now, and those each earlier
     /// attempt uploaded.
     fn create_sent_digests(&self, id: i64, current: Option<String>) -> Vec<String> {
-        let mut sent = self
+        let mut sent = self.db.create_sent(id).unwrap_or_else(|e| {
-            .create_sent
+            warn!(id, error = %e, "reading what a queued create sent failed");
-            .lock()
+            Vec::new()
-            .get(&id)
+        });
-            .cloned()
-            .unwrap_or_default();
         sent.extend(current);
         sent
     }
@@ -1428,6 +1522,8 @@ impl Core {
         parent: &NodeUid,
         name: &str,
     ) -> Result<NodeUid, Box<dyn std::error::Error>> {
+        self.db
+            .note_create_target(op.id, &parent.to_string(), name)?;
         match op.kind == OP_MKDIR {
             true => Ok(self.block_on_bounded(self.drive.create_folder(
                 parent,
@@ -1465,12 +1561,7 @@ impl Core {
         // The call may make the file and lose its answer, and the op may hold
         // a newer blob by the time it is retried (`Core::adoptable_twin`).
         let sha = staged_sha1(Path::new(blob))?;
-        let mut sent = self.create_sent.lock();
+        self.db.note_create_sent(op.id, &sha)?;
-        let digests = sent.entry(op.id).or_default();
-        if !digests.contains(&sha) {
-            digests.push(sha);
-        }
-        drop(sent);
         let thumbnails = self.upload_thumbnails(Path::new(blob), name);
         let guard = self
             .transfers

crates/pdfs-fuse/src/lib.rs

@@ -572,10 +572,6 @@ struct Core {
     /// Drive holds that name until the create lands and a rename or trash
     /// queued behind it frees it again.
     creating: Arc<Mutex<HashMap<i64, (NodeUid, String)>>>,
-    /// The SHA-1 of each blob an attempt at a queued create uploaded, by op
-    /// id. An attempt whose answer was lost may have made the file with bytes
-    /// a later write has since replaced in the op (`docs/BUGS.md` B156).
-    create_sent: Arc<Mutex<HashMap<i64, Vec<String>>>>,
     /// The SHA-1 of the last revision this daemon uploaded to a file and could
     /// not read back, so its revision id is not known (`docs/BUGS.md` B157).
     sealed_unread: Arc<Mutex<HashMap<NodeUid, String>>>,

crates/pdfs-fuse/src/mount.rs

@@ -434,7 +434,6 @@ pub(crate) fn mount_with(
         pending: Arc::new(Mutex::new(HashMap::new())),
         hidden: Arc::new(Mutex::new(HashSet::new())),
         creating: Arc::new(Mutex::new(HashMap::new())),
-        create_sent: Arc::new(Mutex::new(HashMap::new())),
         sealed_unread: Arc::new(Mutex::new(HashMap::new())),
         landed: Arc::new(Mutex::new(HashMap::new())),
         unadopted: Arc::new(Mutex::new(HashMap::new())),

crates/pdfs-fuse/src/sim/daemon.rs

@@ -1290,6 +1290,47 @@ mod tests {
         let _ = std::fs::remove_dir_all(&dir);
     }
 
+    #[test]
+    #[ignore = "mounts FUSE: run with `cargo test -p pdfs-fuse sim:: -- --ignored`"]
+    fn a_write_after_a_restart_over_a_create_never_read_back_is_not_a_conflict() {
+        // The create landed and the link went down before it was read back,
+        // then the daemon restarted. What was to read it back did not outlive
+        // the restart, so the next write was based on the file as it was made
+        // here and landed as a conflict copy of it (B182).
+        let drive = FakeDrive::new();
+        let dir = scratch("unadopted-restart");
+        let daemon = Daemon::start(&dir, drive.client(1, Faults::lan())).unwrap();
+        let file = |bytes: &[u8]| Some(Entry::File(Arc::new(bytes.to_vec())));
+        let path = daemon.mountpoint.join("f.txt");
+
+        // Drive stamps a create with the second it arrives, so it has to land
+        // a second after the file was made here for the two to differ.
+        daemon.client.set_online(false);
+        std::fs::write(&path, b"first").unwrap();
+        std::thread::sleep(Duration::from_millis(1100));
+        daemon.client.drop_link_after_next_create();
+        daemon.client.set_online(true);
+        assert!(wait_until(Duration::from_secs(30), || {
+            drive.tree().get("f.txt").cloned() == file(b"first")
+                && daemon.pending().is_ok_and(|items| items.is_empty())
+        }));
+        let daemon = daemon.restart().unwrap();
+        let path = daemon.mountpoint.join("f.txt");
+        write_at(&path, 5, b" and second").unwrap();
+        daemon.client.set_online(true);
+        assert!(
+            wait_until(Duration::from_secs(30), || {
+                drive.tree().get("f.txt").cloned() == file(b"first and second")
+            }),
+            "{:?}",
+            drive.tree()
+        );
+        assert_eq!(drive.tree().len(), 1, "{:?}", drive.tree());
+
+        assert!(matches!(daemon.stop(), Ok(MountOutcome::Shutdown)));
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+
     #[test]
     #[ignore = "mounts FUSE: run with `cargo test -p pdfs-fuse sim:: -- --ignored`"]
     fn a_create_whose_answer_was_lost_lands_once() {
@@ -1384,6 +1425,126 @@ mod tests {
         let _ = std::fs::remove_dir_all(&dir);
     }
 
+    #[test]
+    #[ignore = "mounts FUSE: run with `cargo test -p pdfs-fuse sim:: -- --ignored`"]
+    fn a_file_deleted_after_a_restart_leaves_drive_with_what_its_create_sent() {
+        // Drive made the file with the bytes its create sent, but the answer
+        // was lost, and the file was written again. After a restart the file
+        // was deleted. The trash knew only the bytes the file held last, which
+        // were not the ones Drive had, so it left that file on Drive (B183).
+        let drive = FakeDrive::new();
+        let dir = scratch("lost-create-reply-restart-deleted");
+        let daemon = Daemon::start(&dir, drive.client(1, Faults::lan())).unwrap();
+        let path = daemon.mountpoint.join("f.txt");
+
+        let held = daemon.client.hold_reply_to_next_create();
+        daemon.client.lose_reply_to_next_create();
+        std::fs::write(&path, b"first").unwrap();
+        assert!(wait_until(Duration::from_secs(30), || held.reached()));
+        std::fs::write(&path, b"second").unwrap();
+        drop(held);
+        assert!(wait_until(Duration::from_secs(30), || {
+            daemon
+                .pending()
+                .is_ok_and(|items| items.iter().any(|item| item.attempts > 0))
+        }));
+        daemon.client.set_online(false);
+        let daemon = daemon.restart().unwrap();
+        let path = daemon.mountpoint.join("f.txt");
+        std::fs::remove_file(&path).unwrap();
+        daemon.client.set_online(true);
+        assert!(
+            wait_until(Duration::from_secs(30), || {
+                daemon.pending().is_ok_and(|items| items.is_empty()) && drive.tree().is_empty()
+            }),
+            "{:?} {:?}",
+            daemon.pending(),
+            drive.tree()
+        );
+        assert!(!path.exists());
+
+        assert!(matches!(daemon.stop(), Ok(MountOutcome::Shutdown)));
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+
+    #[test]
+    #[ignore = "mounts FUSE: run with `cargo test -p pdfs-fuse sim:: -- --ignored`"]
+    fn a_file_renamed_after_its_create_lost_its_answer_lands_once() {
+        // Drive made the file, but the answer was lost, and the file was
+        // renamed before the retry. The retry looked for the file Drive made
+        // under the new name only, and made a second one there (B184).
+        let drive = FakeDrive::new();
+        let dir = scratch("lost-create-reply-renamed");
+        let daemon = Daemon::start(&dir, drive.client(1, Faults::lan())).unwrap();
+        let file = |bytes: &[u8]| Some(Entry::File(Arc::new(bytes.to_vec())));
+
+        daemon.client.lose_reply_to_next_create();
+        std::fs::write(daemon.mountpoint.join("f.txt"), b"first").unwrap();
+        assert!(wait_until(Duration::from_secs(30), || {
+            daemon
+                .pending()
+                .is_ok_and(|items| items.iter().any(|item| item.attempts > 0))
+        }));
+        daemon.client.set_online(false);
+        std::fs::rename(
+            daemon.mountpoint.join("f.txt"),
+            daemon.mountpoint.join("g.txt"),
+        )
+        .unwrap();
+        daemon.client.set_online(true);
+        assert!(
+            wait_until(Duration::from_secs(30), || {
+                daemon.pending().is_ok_and(|items| items.is_empty())
+                    && drive.tree().get("g.txt").cloned() == file(b"first")
+            }),
+            "{:?} {:?}",
+            daemon.pending(),
+            drive.tree()
+        );
+        assert_eq!(drive.tree().len(), 1, "{:?}", drive.tree());
+
+        assert!(matches!(daemon.stop(), Ok(MountOutcome::Shutdown)));
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+
+    #[test]
+    #[ignore = "mounts FUSE: run with `cargo test -p pdfs-fuse sim:: -- --ignored`"]
+    fn a_file_renamed_and_deleted_after_its_create_lost_its_answer_leaves_nothing() {
+        // As above, and the file was deleted after the rename. The trash
+        // looked for the file Drive made under the last name only, and left
+        // it on Drive under the first (B184).
+        let drive = FakeDrive::new();
+        let dir = scratch("lost-create-reply-renamed-deleted");
+        let daemon = Daemon::start(&dir, drive.client(1, Faults::lan())).unwrap();
+
+        daemon.client.lose_reply_to_next_create();
+        std::fs::write(daemon.mountpoint.join("f.txt"), b"first").unwrap();
+        assert!(wait_until(Duration::from_secs(30), || {
+            daemon
+                .pending()
+                .is_ok_and(|items| items.iter().any(|item| item.attempts > 0))
+        }));
+        daemon.client.set_online(false);
+        std::fs::rename(
+            daemon.mountpoint.join("f.txt"),
+            daemon.mountpoint.join("g.txt"),
+        )
+        .unwrap();
+        std::fs::remove_file(daemon.mountpoint.join("g.txt")).unwrap();
+        daemon.client.set_online(true);
+        assert!(
+            wait_until(Duration::from_secs(30), || {
+                daemon.pending().is_ok_and(|items| items.is_empty()) && drive.tree().is_empty()
+            }),
+            "{:?} {:?}",
+            daemon.pending(),
+            drive.tree()
+        );
+
+        assert!(matches!(daemon.stop(), Ok(MountOutcome::Shutdown)));
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+
     #[test]
     #[ignore = "mounts FUSE: run with `cargo test -p pdfs-fuse sim:: -- --ignored`"]
     fn a_file_made_under_the_name_of_a_create_on_the_wire_waits_for_it() {

docs/BUGS.md

@@ -12,6 +12,71 @@ Conventions:
 
 ---
 
+## B184 — A file renamed after its create lost its answer stays on Drive under its old name
+
+**Status:** Fixed (unverified).
+**Found:** 2026-10-05, replaying `one_client_on_a_flaky_link` seed 11 with the B183 fix. `a.txt`'s
+create reached Drive and lost its answer. The file was then rewritten, renamed to `b.txt` and
+then `d`, and deleted. Drive kept `a.txt` with the bytes the create had sent.
+
+**Where:** `Core::withdrawn_twin` and `Core::create_local_node` in
+`crates/pdfs-fuse/src/drain.rs`.
+
+**Cause.** A rename while a create waits rewrites the op's parent and name, but the node a lost
+answer left stays where the create was sent. The trash a delete made of the create, and the
+create's retry, looked for that node under the op's last name only. The trash found nothing
+and left the file on Drive. The retry made a second file under the new name, so Drive had both.
+
+**Fix.** Each attempt records the parent and name it is sent under on the op's row, in a new
+`pending_op.sent_to` column (schema 42). The trash looks for its node under each of them. A
+retry first adopts a node an earlier attempt made under another name, and the rename to the
+current name follows it. Sim tests `a_file_renamed_after_its_create_lost_its_answer_lands_once`
+and `a_file_renamed_and_deleted_after_its_create_lost_its_answer_leaves_nothing`, and
+`what_a_create_sent_outlives_a_newer_blob_its_trash_and_a_restart` in
+`crates/pdfs-core/src/db/tests.rs`.
+
+## B183 — A file deleted after a restart stays on Drive when its create lost its answer
+
+**Status:** Fixed (unverified).
+**Found:** 2026-10-05, by the known-bug confirmation run (`PDFS_SIM_KNOWN=fail`) on da013c2:
+`one_client_on_a_flaky_link` seed 11 failed every replay. `d` was made, written again while
+its create was on the wire, and deleted after several restarts. Drive kept a `d` with the bytes
+the create had sent, and the next file renamed to `d` landed as `d (sync-conflict …)`.
+
+**Where:** `Core::withdrawn_twin` and `Core::adoptable_twin` in `crates/pdfs-fuse/src/drain.rs`.
+
+**Cause.** The trash a delete makes of a create that was sent looks for a node Drive made with
+the bytes the create sent (B151). Those were kept in memory only (B156). After a restart the
+trash knew only the blob the op held last, which a later write had replaced. It found no node of
+its own and left the file on Drive. A create retried after a restart had the same gap and forked
+a conflict copy.
+
+**Fix.** Each attempt records the SHA-1 of the blob it sends on the op's row, in a new
+`pending_op.sent_sha1` column (schema 42). The row keeps it through a newer blob, through the
+turn into a trash and across restarts. The in-memory map is gone. Tests
+`what_a_create_sent_outlives_a_newer_blob_its_trash_and_a_restart` in
+`crates/pdfs-core/src/db/tests.rs`, and sim test
+`a_file_deleted_after_a_restart_leaves_drive_with_what_its_create_sent`.
+
+## B182 — A write after a restart over a create never read back lands as a conflict copy
+
+**Status:** Fixed (unverified).
+**Found:** 2026-10-05, by the known-bug confirmation run on da013c2:
+`one_client_on_a_flaky_link` seed 6 failed two of three replays. `y/x/c.bin`'s create landed and
+the link went down before it was read back. The daemon restarted, and a write to the file landed
+as `c (sync-conflict …).bin`, with Drive keeping the old bytes.
+
+**Where:** `Core::finish_adoption` in `crates/pdfs-fuse/src/drain.rs`.
+
+**Cause.** A create that landed but was not read back is adopted before the next op for the
+node is sent (B143). The node is remembered for that in memory, and a restart forgot it. The
+tree then still had the file as it was made here, with no revision of Drive's, and a write was
+based on that. The drain found Drive's revision in its way and took it for someone else's.
+
+**Fix.** A queued write over a file the tree holds without a revision id is adopted first, as
+B143 does, after a restart too. Sim test
+`a_write_after_a_restart_over_a_create_never_read_back_is_not_a_conflict`.
+
 ## B181 — The app shows a queued edit kept as a conflict copy in English
 
 **Status:** Fixed.
@@ -673,8 +738,7 @@ the op held at the retry was hashed. The twin held the bytes an earlier attempt
 
 **Fix.** Each attempt records the SHA-1 of the blob it uploads, by op id, in memory. A twin holding
 any of them is ours. One holding older bytes than the op's is adopted, and the op's blob goes up
-as a revision of it. After a restart such a twin still forks a conflict copy, which loses
+as a revision of it. Kept on the op's row since B183, so a restart no longer forgets them. Sim test `a_file_written_again_after_its_create_lost_its_answer_lands_once`; the fake
-nothing. Sim test `a_file_written_again_after_its_create_lost_its_answer_lands_once`; the fake
 Drive's held and lost create answers now combine.
 
 ## B155 — A write whose upload lost its answer lands as a conflict copy of itself

docs/CHANGELOG.md

@@ -11,9 +11,10 @@ scratch (user data in `staging/` and `recovery/` is never touched by this).
 
 ## [Unreleased]
 
-Schema: **41**. The queue table is rebuilt so an id is never given out twice; queued changes
+Schema: **42**. The queue table is rebuilt so an id is never given out twice; queued changes
-are kept. Queued changes get a column for the reason Drive refused them, and name their file
+are kept. Queued changes get a column for the reason Drive refused them, and two for the
-and folder by local id as well as by Drive's id. Stored files and folders name their parent
+content a new file was sent with and where, and name their file and folder by local id as well
+as by Drive's id. Stored files and folders name their parent
 folder by local id too. A file or folder made on this machine that Drive does not have yet goes
 by its local id, in the queue, the stored tree and pins alike.
 
@@ -96,6 +97,14 @@ by its local id, in the queue, the stored tree and pins alike.
 - **The daemon no longer hangs when a file being read changes on Drive** (B130).
 - **The Activity page says in your language why an edit was kept as a conflicting copy**,
   instead of the daemon's English (B181).
+- **A file written after a restart no longer lands as a conflict copy of itself** when the
+  connection dropped just as it first reached Drive (B182).
+- **A file deleted after a restart is removed from Drive too** when the connection lost Drive's
+  answer to its upload and the file was written again meanwhile. It no longer stayed there and
+  turned the next file under its name into a conflict copy (B183).
+- **A file renamed while the connection lost Drive's answer to its upload lands once.** It no
+  longer stayed on Drive under its old name too, or, when deleted, under its old name alone
+  (B184).
 - **A file renamed just as its upload finished keeps its new name.** The mount no longer showed
   the old name, and a second rename no longer failed with "No such file or directory" (B152,
   B172).

docs/MILESTONE-3.0.0-STATUS.md

@@ -9,7 +9,7 @@ Tracks `docs/MILESTONE-3.0.0.md`. Update it with every commit that moves a phase
 - **Phase 2**: local ids, inode = local id, one tree for every mount (2.10.0).
 - **Phase 3, first part**: mkdir, create, rename, unlink and rmdir on the mount are recorded
   locally and sent from the queue; `"local_first": false` brings back the old path (5e0f30c).
-- Fixes from the simulation and account runs since then: B125, B127, B130 and B141 to B181.
+- Fixes from the simulation and account runs since then: B125, B127, B130 and B141 to B184.
 - **Sync issues, first cut** (§6.2): refusals sorted into issues, shown at once in the app,
   tray and CLI; Export button and `pdfs sync export`; `pdfs sync issues`; Discard button and
   `pdfs sync discard`; a warning badge on the node in the file browser.
@@ -102,7 +102,12 @@ Tracks `docs/MILESTONE-3.0.0.md`. Update it with every commit that moves a phase
          which a68dae9 (not yet pushed then) also takes off the runtime. CI now keeps a hung
          run's stacks.
    - [ ] Known-bug list in `sim/run.rs` is empty: none open; B125, B127, B129, B130 and B132
-         fixed but not yet confirmed by the runs.
+         fixed but not yet confirmed by the runs. The confirmation run (`PDFS_SIM_KNOWN=fail`)
+         on da013c2 failed `one_client_on_a_flaky_link` seeds 6 and 11 on two new bugs, both
+         after restarts: a write over a create never read back (B182) and a trash that forgot
+         what its create sent (B183). Replays with those fixed found a third: a create renamed
+         after it lost its answer (B184). All three fixed; no known-bug matcher fired. It was
+         stopped before `three_clients`.
    - [ ] A clean account run. The one on a6f36d5 (2026-10-04, LAN: 427 passed, 12 failed)
          failed B113 and B121 on every on-demand mount, two mirror-to-mirror move cases, and
          the journal check. B121 was a bug (B149, fixed); B113's case still expected creates